I spent 23 years in education administration watching technology arrive and asking the question everyone forgot: not what can it do, but why do we need it here, and what happens when it goes wrong.AI was no different — except the gap it exposed was bigger than anything I had seen. No governance. No accountability layer. No structure around the decision boundary.This is the lab notebook. The posts below document what I found, what I built, and why the answer was never the model.
Building Cognition: My Journey from Chatbot to Structure
Not what AI can do — but why it keeps failing, and what I built to fix the governance gap nobody was naming.
The Perfect Storm Nobody Named
AI made work faster. Everything else got harder. Here is what was actually happening, why every attempted fix failed, and what the data says about the cost of having no governance at the front end.
The Greenfield Test: Can You Regrow a System From Cards?
29 cards, one eleven-word prompt, four phases. What happens when you harden intent before writing a single line of code.
The Reliability Wall: Why We Must Onboard AI Like Workers, Not Tools
Vector Institute and Unilever let an agent write its own SQL against production data. It failed in familiar ways. Their fix — less freedom, not a smarter model — points at a bigger gap: the difference between a workflow gate and a governance layer.